Mirror effect in order and associative information: Role of response strategies.

Classes of stimuli that are relatively easy to classify in recognition as old when old also appear to be relatively easy to classify as new when new. Five experiments are reported that extend this mirror effect to tests requiring retention of order and associative information. These experiments tested memory for words and nonwords; across experiments, the mirror effect was found on both yes--no and forced-choice tests. In addition, a new account of the mirror effect is proposed, one which suggests that this pattern results from participants' attempts to distribute responses equally across stimulus classes on tests. Support from this account came from additional experiments in which the mirror effect was eliminated when participants were asked to refrain from using positive responses as guesses. Numerous studies have been performed in which recognition memory has been tested for study lists containing two distinct sets of stimuli. The typical finding is that if one class is easier to classify as old when old than the other, then that class will also be easier to classify as new when new. On a yes-no recognition test, this pattern is shown when the stimulus class that has the higher hit rate also has a lower false-alarm rate than the other class. On a forced-choice recognition test, the equivalent pattern would be shown when participants are most accurate on test pairs containing a target and a distractor from the same class, least accurate on test pairs containing a target and a distractor from the other class, and intermediate in accuracy on the pairs containing one stimulus each from the two classes. Results such as these have been called the mirror effect and have been found in many recognition studies, such as comparisons of words with nonwords and low-frequency words with high-frequency words (Glanzer & Adams, 1985, 1990). The generality of the mirror effect across stimuli in recognition has been clearly demonstrated. The purpose of Experiments 1-5 was to demonstrate its generality across memory tests. It has been shown that the mirror effect can be found in frequency discrimination (Greene & Thapar, 1994). Frequency discrimination, like recognition, is a test of item information, and evidence suggests that frequency judgments and recognition are performed using similar processes (Hintzman & Curran, 1994). The first five experiments reported here extend the mirror effect to measures of order and associative information. All current accounts of the mirror effect have been intended to explain its occurrence on recognition tests. If frequency discrimination is performed using the same processes as recognition, such theories may be able to be extended